A flurry of blogs have been written over the past week by Sepaton, Mr. Backup, Veeam, etc. about the survivability of companies like Veeam that have developed specialty products for addressing VMware backup challenges. Here’s a list of a few…
The back and forth basically got a head of steam because of comparisons to VM backup boutique products to CDP — Continuous Data Protection technology — products of a few years back. There was also a lot of hyperbolistic characterization of backup software stalwarts as “dinosaurs.”
The issue in a nutshell hinges, first and foremost, on the question of whether VMware will become as predominant as its marketeers think. Clearly, folks who are writing applications to capitalize on the current VMware craze expect the “product that is also a movement” to achieve its ambitious market penetration goals. In shops that are already deploying the hypervisor, backup, among other aspects of computing, has been negatively impacted by the technology. The claim of the boutique VM backup products is that they have the solution that the Symantecs, CAs and Tivolis can’t provide…yet.
The response of the pundits seems to be to question whether — if and when the hypervisor revolution really happens — a couple of upstarts who have some market share will thrive, or the big guys will simply buy them out to get their IP. The commenters seem to be arguing the latter position, which sounds reasonable to me.
I frankly don’t understand all the fuss. In talking to several server virtualized shops recently — including a Baptist ministry in South Florida, a mega hosting/co-lo service provider, and a medical staffing company — all of them seem to have this VM backup thing down. They have virtualized their storage infrastructure with DataCore Software. They can protect data for specific VMs or for entire hosting environments or on a guest basis even through a number of replication process alternatives. They can do CDP for any VM requiring periodic restore points, or they can mirror between two or more locations asynchronously, or they can copy after write to a stand of disks set up as a VTL so they can backup to tape. Easy peasy.
The video clips I shot with these users will be on-line shortly at the C-4 Project, with an abbreviated version being shown on DataCore’s own website. If you were at VMworld (I wasn’t) you could have seen the interviews on a DVD we cut for DataCore to use at the show.
When I get together with FalconStor next week in Long Island, I will ask about their approach and get that on video as well. Stay tuned.
Truth be told, I don’t get the debate at all — except for the commentary of the boutique software guy, who is trying to get some street cred for his stuff.
As for the comparison with CDP raised by some of the insightful critics, I think a point is being missed that makes the comparison specious. They cite the experience of Revivio as a boutique CDP backup product that came to market too early and was subsequently gobbled up by one brand name vendor or another.
I guess that is one way of looking at it. Kirby Wadsworth, who helmed the company and is now at F5, could probably tell the real tale. My take on Revivio was that it was a shot across EMC’s bow. EMC had long offered its own Point in Time Mirror Split Imaging solution, which involved pricey software run on the array, that used the most expensive disk in your infrastructure to make mirrors of the most expensive disk in your infrastructure.
Revivio offered the same functionality but let you use an externalized stand of inexpensive disk, Joe’s JBODs if you preferred, to make your CDP PIT Split copies. The problem, I thought, was that folks who already drank EMC’s Kool Aid were disinclined to add in third party wares in any case.
The other problem with CDP was that database software companies were adding crash recovery features to their software that negated to some degree the perceived need for PIT copying in the first place.
And finally (and this is just my own spit-balling) I was hoping that folks were waking up to the fact that PIT mirroring on databases was a fool’s errand. The fact that most database corruption events weren’t detected until a substantial period of time after they had occurred often meant that all of the timestamped copies of data that you had made were already corrupt. Better have a tape backup handy.
Now, that’s CDP of a few years ago, when a gaggle of companies got VC funding to try to out-EMC EMC. The term itself was coopted by other players in the storage market until it was so watered down that it became meaningless. Continuous data protection might just as well apply to time interval snapshots or incremental tape backups. Both provide data protection at time-defined intervals, capturing data state changes. The questions of efficacy came down to issues of frequency and granularity.
The core issues of VM backup have to do with two things: first, where is the guest and its data-to-be-backed-up actually located in a world where guests move around at will? Storage virtualization seems to make short work of this issue since they control the data and expose it to replication services per policy. Second, can VM backup be run efficiently using agents? VMware seems to be working this issue on its own, providing a standard API for backup ISVs to leverage as well as location where data (pointers?) can be stored so they can be accessed more readily by backup apps. I suspect that all ISVs will shortly leverage this API.
So, much ado about nothing…from everyone’s perspective except boutique backup software vendors who want to sell into the VMware user caucus and perhaps VMware itself, which continues to try to represent its Jenga! game for application stacking as more highly available and resilient than other hosting models.